Self-Selection and the Earnings of Immigrants: Comment
评论Borjas关于美国移民自我选择的论文，指出其样本选择偏差导致结果不一致；通过扩大样本和考虑选择性移民，发现美国吸引高技能移民，且移民收入变化受选择性移民影响。
In a recent article in this Review (Borias, 1987), George Borjas uses a standard model of self-selection to demonstrate that immigrants in the United States may not necessarily be positively selected, as is commonly assumed. Borjas' empirical results, based on U.S. Census micro data merged with country-of-origin characteristics for 41 sending countries, however, do not provide clear or consistent results with respect to the immigration selection issue, although they do confirm earlier findings that the characteristics of the countries of origin of U.S. immigrants explain a substantial proportion of the differences in their economic status (Jasso and Rosenzweig, 1986b).' The purpose of this note is to show that Borjas' results are themselves subject to biases due to additional processes of self-selection. In particular, his inconclusive findings with respect to immigration selectivity arise from the use of a choice-based sample; enlargement of his highly selective sample of countries, based on the number of foreign-born in the United States, yields results that appear to conform much more closely and consistently to conventional wisdom, namely that the United States appears to attract those persons with above-average skills. We also provide evidence that changes in the earnings of aggregate immigrant entry cohorts, used by Borjas to measure the assimilation of immigrants, reflect another evidently important selection process, namely selective emigration. The effects of country characteristics on the wages of the foreign-born at entry (immigration selectivity) are thus not comparable to the effects of such variables on the changes in the characteristics of an immigrant cohort over time (emigration selectivity). The effects via immigration self-selection of country-of-origin characteristics on the Q of immigrants who migrate to a destination country is neatly summarized by Borjas in his equation (12). With unobservables determining earnings normally distributed, the expected quality of immigrants in the United States is given by